Astructure’s influence on tunnelling-
induced ground movements
D. M. Potts, BSc(Eng), PhD(Cantab), MICE, CEng, and T. I. Addenbrooke,
BEng, ACGI, MSc, DIC
The influence of tunnelling on buildings has
become an important issue. The problem is
an interactive one; not only do tunnelling
settlements affect existing structures, but
existing structures affect tunnel-induced soil
movements. In this paper the results of a
parametric study of the influence of an
existing structure on the ground movements,
due to tunnelling are presented. The width
of the structure, its bending and axial
stiffness, its position relative to the tunnel
and the depth of the tunnel are considered.
‘The interaction is shown by reference to
commonly-used building damage para-
meters, namely deflection ratio and max-
imum horizontal strain. By introducing
relative stiffness parameters which combine
the bending and axial stiffness of the
‘structure with its width and the
the soil, design curves are established.
These give a guide as to the likely mod-
ification to the greenfield settlement trough
Introduction
‘The assessment of the influence of tunnelling on
buildings and other structures has become an
important and costly environmental issue. For
example, a large proportion of the petitions
against the Jubilee Line Extension in London
were settlement-related. There is, therefore, and
hhas been for some time, a pressing need for
research on the performance of structures
subjected to tunnelling-induced settlements. The
problem is, however, an interactive one ~ not
only do tunnelling settlements affect existing
‘structures, but existing structures affect tunnel
ling settlements.
2 Current design practice depends on
empirical methods for the prediction of tunnel
ling-induced ground movements. The methods
are based on historical data from greenfield
sites.'” If the effect of ground movements on a
surface structure is to be assessed, then the
building is assumed to be infinitely flexible,
and to follow the greenfield settlement profile.
‘The translations, rotations, strains and defor-
mations so predicted are then compared with
limiting criteria to estimate the likely damage to
the building, * It is felt that this approach,
based on a greenfield settlement profile, and
making no allowance for the stifiness of the
structure, can be improved to account more
accurately for urban environments where exi
ing surface structures modify the ground
‘movements.
3. The study presented in this paper ad-
dresses one aspect of this problem, namely the
influence of an existing surface structure on the
ground movements due to tunnelling. A series
‘of numerical analyses of tunnel construction in
the greenfield scenario, and beneath-surface
beams of varying stifiness (representing struc-
tures) is used to assess the differences between
greenfield ground movements and those modi-
fied by the structure. Building damage para
meters (deflection ratio and horizontal strain)
are considered for the modified ground move-
‘ments with respect to the greenfield values. A
unifying framework is then developed to ac-
‘count for the relative soil/structure stiffness
when predicting levels of building damage. A
design approach is thereby proposed which
gives an improved method of assessment of the
magnitude of building damage in response to
tunnelling-induced ground movements.
Parametric study
4. ‘The geometry of the problem under
investigation is shown in Fig. 1. The excavated
tunnel diameter was fixed at D = 4-146m and
the depth from the soil surface to the tunnel
axis, Z, was either 20m or 34m These values
are typical for 2 London Underground running
tunnel. A beam of width, B, resting on the
ground surface with its centre at an offset
distance, ¢, from the tunnel centreline was used
to represent the effect of a surface structure.
‘The main variables considered in the parametric
study were the axial and bending stiffness EA
‘and EI (where Eis the Young's modulus, 4 the
cross-sectional area and I the second moment
of area of the beam) along with the beam width,
B, and its eccentricity with respect to the tunnel
centreline, e
5. The sol profile was assumed to be
London Clay and was represented by a non-
linear clastic plastic constitutive model. The
model described by Jardine et al.’ was used to
represent the nonlinear elastic prevyield beha-
Proc. Instn
Ci. Engrs
Geotech. Eman,
1997, 125, Apr,
109-125
Paper 11048
Written discussion
loses 16 June 1997
Manuscript recived
122 November 1995; revised
‘manuscrit accepted
8 May 1996
D. M, Potts
Professor of
Analstical Soil
‘Mechanics,
Department of Civil
Engineering.
Imperial College of
Science, Technology
and Medicine,
London
TI, Addenbrocke,
Lecturer,
Department of Civil
Engineering,
Imperial College of
Science, Technology
‘and Medicine,
LondonPOTTS AND
ADDENBROOKE
110
viour, and a Mohr-Coulomb yield surface and
plastic potential were used to model the plastic
behaviour. Appendix 1 gives more details of this
‘model and the values of the input parameters.
‘The initial stress state in the ground was
assumed to be controlled by a saturated bull
unit weight of 20kN/m}, a hydrostatic pore
water pressure profile with a water table located
2m below ground surface and a coefficient of
earth pressure at rest, Ky = 15. Only a short-
term response was investigated and therefore
the soil was assumed to behave undrained. This
was modelled in the analysis by assigning a
high bulk modulus to the pore water (= 100
times K,, where Ky is the effective bulk
modulus of the soil skeleton).
6, The surface beam used to represent an
overlying structure was assumed to be elast
and its interface with the soil to be rough. For
‘many of the analyses the axial and bending
stif ness were varied independently over an
extreme range. While this is necessary from an
academic point of view so that the extreme
limits of behaviour can be investigated, it does:
‘sometimes result in an unrealistic combination
of axial and bending stiffness. Additional ana-
lyses were therefore performed in which the
stiffness values adopted for the beam were
based on a single slab or a 1, 3, or Sstorey.
structure. The single slab was assumed to be of
150: mm thick concrete having a Young's mod.
ulus £ = 23 x 10° kN/m?, the 5-storey structure
hhad six of these slabs with a 3.4m vertical
spacing, the 3storey structure four slabs, and
the I-storey structure two slabs, The calculation
of equivalent £, A, and I values of the surface
beam in these cases assumed axial straining
along each structure's full height to give axial
stiffness, and employed the parallel axis theo:
rem to define each structure's stifiness against
bending about its neutral axis (Appendix 2)
‘This can be considered to be an overestimate of
building stiffness, as only a rigidly framed
structure would approach such modes of
deformation. An alternative assumption is to
obiain the bending stiffness by summing the
independent EI values for each storey.” This
implies that the walls and columns transfer the
same deformed shape to each storey.
7. In all cases no vertical loads were applied
to the surface beam, therefore the dead and live
loading of the structure are not modelled. Such
loading is likely to affect both the shear
strength and stiffness of the adjacent soil.
Although this is a simplification such loading is
ly to vary considerably in practice and to
have included it as a variable in the present
study would have greatly increased the com-
plexity of the investigation.
8 To account for the stiffness of both the
beam (structure) and the soil the following two
measures of relative stiffness are introduced
and will be used subsequently when the results
poooooogga
Odfjoooo qa
from the analyses are presented. The relative
bending stifiness, p*, and relative axial stiffness,
a*, are defined as
~_ EI EA
Ei BA
where HY is half the width of the beam (= B/2)
and B, i a representative sil stifness, The
expression for p* is similar to that used by
Fraser and Wardle’ and Potts and Bond,® and
that for a ig similar to that used by Boseardin
and Cording.' It should be noted that for the
Present investigation, which involves plane strain
conditions, a* becomes dimensionless while p*
has dimensions of m-!. The value of , adopted
in the present work is the secant stifiness that
would be obtained at 0-01% axial stain in a
triaxial compression test performed on a sample
retrieved from a depth of Z/2. This was chosen
as itis a measure that could be obtained from a
site investigation.’ For the soil profile used in
the present investigation, E, for the tunnels with
a depth, Z, of 20m and 34m is 103 MPa and
163 MPa respectively. The value of, increases
approximately linearly with Z.
9. The building damage parameters adoy
are deflection ratio and horizontal strain."
Deflection ratios for both sagging, DRug. and
hogging, DRog, are defined, see Fig. 2. a
e a
Detection ratios:
OP,
Fig. 1. Problem
geometry
Fig, 2. Definition of
deflection ratiospoint of inflection of the surface settlement
trough occurs below the beam then it separates
the zones of sagging and hogging. In the
analysis this point was determined by interro
gating the surface settlement troughs to locate
the point at which the rate of change of slope of
the trough changed sign. The horizontal strain,
2, is obtained directly from the computer
‘output and is the maximum horizontal strain of
the neutral axis of the beam and therefore of
the structure the beam represents, By referring
the strain to the neutral axis any effects of
bending are eliminated
10. When presenting the results from ana
Iyses with a surface structure, comparisons will
be made with greenfield predictions. It is
therefore convenient to define the following
modification factors for deflection ratio
where ef. and of, are the maximum horizon
compressive and tensile strains of the ground
surface for that portion of the greenfield
settlement trough which lies directly beneath the
structure.
Finite clement analy
11. The finite element program ICFEP (Im-
perial College Finite Element Program) was
used to carry out the analyses reported here, A
typical finite clement mesh is shown in Fig. 3
For beams with zero eccentricity, advantage was
taken of geometric symmetry and consequently
only half of the problem was analysed, as was
cevident from the mesh shown in Fig. 8. For
eccentric beams such symmetry does not exist
and the complete problem was analysed. Fight
noded plane strain isoparametric elements with.
reduced integration were used to represent the
soil. Three noded Mindlin beam elements with
selected reduced integration were used to
‘model the surface structures." These elements
simulate bending, axial and shear behaviour of
the beam. Bending behaviour is governed by EI
while the axial and shear behaviour by EA.
Consequently if £4 is small then both the axial
id shear sfifiness will be low.
12, An accelerated modified Newton-Raph-
son scheme with a substepping stress point
algorithm was employed to solve the norlinear
finite element equations."
13. Tunnel excavation was modelled by the
incremental removal of the solid elements
within the tunnel boundary. That is the stresses
that the soil within the tunnel applied to the
‘tunnel boundary were evaluated and then
applied in the reverse direction over several
increments. During this procedure the soil
elements within the tunnel were not included in
the analysis. For each increment of the analysis
the movements of the tunnel boundary were
monitored and used to calculate the volume of
‘STRUCTURE INFLUENCE
ON TUNNELLINGINDUCED.
GROUND MOVEMENTS
co
Fig. 3. Typical finite
element mesh
i
lL